No description available

Simple PyTorch profiler that combines DeepSpeed Flops Profiler and TorchInfo

A tool to parse PyTorch profiler trace files for kernel-level analysis.

A toolkit to parse torch profiler data source and produce spreadsheet

LLM Profiling with DeepSpeed Flops Profiler & Torch Profiler

Example of How to Use the PyTorch Profiler with PyTorch Lightning

A conda-smithy repository for torch-tb-profiler.

TorchProf by torch.profiler

No description available

No description available

A guide showing a little hack to get more info out of the torch profiler.

Runs and collects performance and power profiling data for various MIMD optimized CPU/GPU/NPUs

No description available

No description available

An example of how to use the PyTorch profiler to analyze the performance of a machine learning algorithm.

[WIP] Torch Profiler (Based on Verl Project)

One-stop PyTorch GPU profiling & optimization skill｜Detect memory leaks, speed up training, fix device mismatches, and maximize GPU utilization Works with Cline / Claude / Continue.dev / VSCode AI｜Out-of-the-box｜CLI one-click analysis

This is based on the torch_tb_profiler source code and fix some bugs.

Operation-level profiler for Apple Silicon / MLX. The missing torch.profiler equivalent for the MLX ecosystem.

GPU Training Pipeline Profiler — PyTorch + mixed precision + torch.compile optimization benchmarks

CPU vs GPU performance benchmarking using PyTorch, Torch Profiler, and NVIDIA NVBandwidth

Profile-driven PyTorch inference optimization: 15.65x speedup on ResNet-50 (RTX 4090). Demonstrates torch.profiler, channels_last, FP16, torch.compile.

A performance profiling study of LLM fine-tuning pipelines. Analyzes DataLoader bottlenecks, CUDA synchronization, and torch.compile (Inductor) speedups using PyTorch Profiler and Weights & Biases.

GPU profiling baseline for PyTorch training on Amazon EKS Auto Mode — ResNet-18 on CIFAR-10 with torch.profiler, Terraform-provisioned infrastructure, and Kubernetes Job manifests.